Estimation of Soil Properties by Orbital and Laboratory Reflectance Means and its Relation with Soil Classification

Wet chemistry methods to extract soil properties such as Fe2O3, TiO2, MnO and clay are cost effective, time consuming and environmental polluter. Moreover, a large set of samples has to be collected for precise spatial mapping. Ordinary surface soil mapping is a problematic method. Accordingly, non destructive technologies, such as remote sens- ing methods can provide important vantages. The objective of the present work was to estimate soil attributes by labora- tory and orbital sensors and compare these results with soil classification. The study area is a 473 ha bare soil field located in the region of Barra Bonita, Brazil. A sampling grid of 100 by 100 m was established and the exact position of each point was georeferenced, and sent to traditional (wet) laboratory analyses. The soil samples reflectance were also acquired by a laboratory sensor using artificial illumination (450 to 2500 nm). Over the same selected ground area reflectance data were extracted from the TM-Landsat-5 image. Prediction equations between the satellite and laboratory reflectance data and the wet chemistry were generated for each attribute. Most of the generated equations presented high and significant R 2 such as for the Fe2O3 with 0.82 for laboratory and 0.67 for the orbital reflectance data. The comparison between reflec- tance estimates and laboratory wet measurements for iron presented 92.2% success for the laboratory and 91.3% for the orbital sensors. The comparison for the texture intervals, showed 65% and 50% success for laboratory and orbital data re- spectively. The iron contents obtained by the sensors allowed to better remotely classify soil classes. Soil extractions to determine these attributes can be substitute by spectral reflectance models based on the present methodology.